OneTrack.AI bolts AI cameras onto forklifts to catch what the spreadsheet never does - how product actually moves, and how people actually get hurt.
And it turns out those are two very different datasets - which is, more or less, the entire business.
Here is a fact about warehouses that is either obvious or deeply strange, depending on how much time you have spent in one: the software running a modern distribution center knows, with great confidence, what is supposed to be happening. It knows that pallet 4471 was scanned at dock 3 at 9:04 a.m. and is now, in the eyes of the database, resting peacefully in aisle G. What the software does not know is that a forklift clipped a rack on the way there, that the driver took a shortcut through a pedestrian lane, or that the top case is crushed. The system tracks transactions. It does not watch.
OneTrack.AI is, at its core, a company built on that gap. Founded in 2017 by Marc Gyongyosi - who was, at the time, doing robotics work tied to BMW and noticing that production lines ran to sub-second precision while the warehouse next door still had people hunting for pallets with binoculars - the company sells a deceptively simple proposition. Put a camera where the work actually happens. Let a computer-vision model watch. Then tell the humans what it saw.
The clever part, and the part that makes this a real business rather than a demo, is where the camera goes. The obvious move in warehouse vision is to hang cameras from the ceiling, which gives you a lovely and largely useless view of the tops of people's heads. OneTrack started with drones, decided drones were a headache, and landed somewhere better: the forklift. The forklift goes everywhere the work goes. It is already moving. Bolt an AI vision sensor to it and you have, essentially, a roving eyewitness on every important machine in the building, capturing the route taken, the way product is handled, and the condition of the load when it leaves.
What you do with that footage is the whole game. OneTrack runs it through deep-learning models on low-cost edge hardware and turns it into three things a warehouse executive genuinely cares about: fewer accidents, more throughput, and cleaner shipments. The safety product flags risky behavior - a near-miss, a pedestrian in the wrong place - and fires a real-time alert plus video evidence, so a supervisor can coach the behavior instead of writing it up after the ambulance leaves. That distinction, prevention versus paperwork, is one the company returns to constantly, and it is a reasonable thing to be smug about.
Figures reported by OneTrack.AI and its customers. Treat vendor-reported results as approximate and directional.
The productivity story is quieter but arguably more durable. OneTrack watches for gap time - the dead minutes where a forklift is idling, a task is stalled, an operator is waiting on something upstream - and the company's stated philosophy is to stack small wins rather than chase a moonshot. A 10 to 15 percent throughput bump does not make headlines, but across a 40-site network like CJ Logistics, an 11 percent units-per-hour increase is the kind of number that pays for a lot of cameras. The quality product, meanwhile, uses the same footage to catch over, short, and damaged shipments before the truck leaves the dock, which is a polite way of saying it prevents the argument that would otherwise happen three days later over the phone.
More recently the company has leaned into the word of the moment: agentic. Its platform now includes AiOn, pitched as an AI operations analyst that reasons over the floor data and the warehouse's existing systems rather than just displaying a dashboard for a human to squint at. This is the fashionable frontier of AI - agents that do things - and it is worth noting that OneTrack is pointing one at a warehouse floor rather than an inbox. Whether AiOn is a genuine autonomous analyst or a very good copilot is the sort of thing that depends heavily on where you draw the line, and reasonable people will draw it in different places. What is not in dispute is that OneTrack has a large, proprietary, and unusually physical dataset to point it at.
The customer roster is the strongest tell. This is not a company selling to a handful of curious pilots. Kellanova, Church & Dwight, Hain Celestial, Ryder, CJ Logistics, ID Logistics, The Wonderful Company, Bigelow Tea, GE Appliances, Holman Logistics - these are large, unglamorous, cost-obsessed operations, and they are exactly the buyers who do not renew software that does not work. That OneTrack pitches itself as live "in days" rather than quarters matters here too, because the enemy of enterprise warehouse software is usually not the technology but the eighteen-month integration nobody finishes.
*GE Appliances figure reflects a reported reduction in claims. All figures are vendor- and customer-reported reductions over stated deployment windows.
Low-cost AI cameras and edge sensors mounted on forklifts and around the floor, capturing routes, handling, and load condition in real time.
Detects risky behavior and near-misses, fires real-time alerts, and delivers video evidence for coaching - targeting 90%+ fewer incidents.
Finds gap time and idle hours, typically lifting throughput 10-15% by turning dead minutes into moved product.
Catches over, short, and damaged (OS&D) issues at the dock, before the truck leaves and the dispute begins.
Monitors forklift-fleet usage to surface idle assets and improve how equipment gets deployed across a site.
An agentic AI operations analyst that reasons over floor data and connected WMS/ERP systems - not just another dashboard.
Marc Gyongyosi, working on robotics tied to BMW, clocks the visibility gap between precision production lines and improvised warehouses.
The drone concept gives way to AI cameras on forklifts. First commercial deployment lands in 2019.
Consumer-goods giants and major 3PLs adopt the platform, citing double-digit productivity and steep safety gains.
OneTrack reframes around agentic AI, launching AiOn and "Agentic AI for Supply Chain."
Clear-cut approaches that repeat and scale.
Field execution over perfectionism.
Compounding gains, not moonshots.
Outcomes over titles; challenge the assumptions as the market moves.
Product walkthroughs, the AiOn agent, and warehouse footage from the company channel.
youtube.com/@OneTrack-AI →Founder Marc Gyongyosi on why vision beats transactions - podcast & interview series.
Listen / watch →Sources: OneTrack.AI, Forbes, Crunchbase, PitchBook, Tracxn, Inbound Logistics, States Logistics. Metrics are vendor- and customer-reported and should be read as approximate.